This invention relates to communications systems in which information is transmitted from a sending station to a receiving station in the form of electrical or electromagnetic information bearing signals over a communication link having a predetermined system frequency bandwidth. More particularly, this invention relates to methods and systems for reducing the actual bandwidth occupied by the information bearing signals to improve the flow of information or to reduce the actual bandwidth requirements of the communication link. Specifically, this invention relates to systems and methods for reducing the actual bandwidth requirement of that class of signals having inherent time division and frequency division multiplexing, such as human speech signals.
It has long been known that the primary intelligibility of human speech (defined as the speech information content as opposed to an individual speaker's identifiable characteristics, such as voice timber and the like) lies in the band from about 1000 to about 3000 HZ, and that human speech is naturally temporally divided into higher frequency components (the consonants occurring in the range from about 1500 to about 3000 HZ and lower frequency components (vowels) occurring in the range from about 0 to about 1500 HZ. Stated differently, human speech may be characterized as information bearing signals having inherent time division and frequency division multiplexing due to the serial nature of speech pronunciation: i.e. a vowel and a consonant cannot occur at the same time. It should be noted that the time division and frequency division multiplexing of human speech is not absolute, i.e. vowel sounds may have frequency portions lying above 1500 HZ while consonants may have frequency components lying below 1500 HZ. However, from a statistical point of view, if significant energy in a particular speech utterance exists below 1500 HZ then the probability of the existence of a very small amount of energy above 1500 HZ during the same time period is close to one. Conversely, if a high percentage of speech energy exists above 1500 HZ, there is a high probability that there will be very little energy below 1500 HZ for that same time period.
Efforts have been made in the past to exploit the above natural characteristics of human speech to reduce the actual bandwidth required to transmit such information from a sending station to a receiving station in order to (a) permit the use of a communication link having a bandwidth less than the normal bandwidth of 3000 HZ, or (b) permit more than one set of speech signals to be transmitted over the same communications link having the normal speech bandwidth of 3000 HZ. In an early system exemplified by the disclosure of U.S. Pat. No. 1,836,824 directed to a telephone communication system, parallel signal conduction paths are provided in the transmitting and receiving portion of a telephone apparatus, with each pair of signal paths having a first path for passing only signals lying in a relatively low frequency range of 250-2250 Hz and a second path having means for passing only higher frequency signals lying in the range from 1250-3250 Hz. Signals in this higher frequency range are converted to equivalent signals lying in the lower frequency pass range of the first path with the output of each signal path being coupled to a common transmitting or receiving unit. The first and second signal paths are operated in a mutually exclusive mode by a switching device which permits the input signal to be coupled to either the lower frequency signal path or the higher frequency signal path, depending on the amplitude of the input signals, it having been experimentally observed that vowel sounds statistically have a greater amplitude than consonant sounds. Systems of this type suffer from the disadvantage that the switching elements inject undesirable noise into the speech signals transmitted or received, which noise predominantly lies within the ordinary frequency band of the communications system, and further from the disadvantage that the switching from the normally active higher frequency signal path to the lower frequency signal path is exclusively dependent upon the amplitude of the speech input signals, so that the occurrence of a consonant sound at a relatively large amplitude causes such relatively high frequency signals to be coupled to the relatively low frequency signal path with resultant signal loss.
A later approach employing a more sophisticated method and system, and exemplified by U.S. Pat. No. 2,726,283, employs a switching device at the output end of parallel signal paths and additional decision making circuitry for measuring the fundamental pitch of the input speech signals to enable time-sharing of the period of the fundamental by causing the low frequency signals to be coupled through the mixer switch during the first half period of the fundamental and the down converted consonant equivalent signals to be coupled through the mixer switch to the output terminal during the second half period of the fundamental. This arrangement suffers from the disadvantage that inherently one-half of the speech signal is lost since the normal temporal separation of the vowel and consonant portions of the speech signal is not relied upon. In addition, this arrangment suffers from the noise injection problem noted above due to the controlled switching that also requires rather complex electrical circuitry which is subject to frequency drift and other distortion introducing characteristics. Further, the circuitry required to measure the fundamental period adds further cost and complexity to the entire system.
Efforts to date to provide a simple, economical method and system for accomplishing the objectives noted above with the signals having inherent time divisions and frequency division multiplexing have not met with success.